Towards Benefiting Both Cloud Users and Service Providers Through Resource Provisioning

Introduction

Cloud computing is one of the emerging trends in technology, and it is not unfamiliar anymore. Most of the businesses these days are moving to the cloud because of its flexibility, scalability, lower maintenance, resiliency, redundancy and especially on-demand service provisioning. Cloud consists of pooled resources, providing services to remote users over the internet. Cloud datacenters situated in various parts of the world, hold numerous high-end resources which are powerful in their performance (Ahuja & Deval, 2018).

There are three primary service layers in the cloud known as Software-as-a-service SaaS), Platform-as-a-service (PaaS), Infrastructure-as-a-service (IaaS). In SaaS, the software lies in the cloud and user can access and use it over the internet (Katzis, 2015). Through PaaS, any web developer or business can build software application and service of their own. So, that provides a platform, developing tools to create software, test and host it in the cloud. In IaaS, users can rent infrastructure, i.e., the hardware of their choice for storage, computation, hosting, etc. The hardware is provided to them are the virtualized parts of the underlying cloud hardware. IaaS subjects to 70% of overall services provided by the cloud, the rest 30% being PaaS and SaaS.

Resource management is the activity through which all the events occurring in the cloud are handled (Singh & Chana, 2015). Various resource management events are shown in Figure 1. Resource management is done in two phases, known as resource provisioning and resource scheduling. Resources provisioning is the process through which suitable resources are discovered and allocated to the workload(s). Whereas, resource scheduling is mapping the resources to the workload and executing them. Resource provisioning is usually performed before resource scheduling. Any errors in the management of resources may cause SLA violations, leading to high penalties, low customer satisfaction and therefore a loss of revenue and customers (Durga, Mohan, Peter & Surya, 2018). The State-of-art resource provisioning techniques lack in balancing the benefits of the clients and the cloud service provider. In particular, automatically identifying suitable resources to the clients’ request is of great concern as it impacts on the service response time and the cost. Grounded on this problem, effective resource provisioning is an important aspect that need to be considered in a cloud environment. The objective of this research is to propose an efficient cluster-based resource provisioning (CB-RP) technique that balances cloud user requirements and CSPs benefits.

Figure 1.

Resource management events

Background

When a user requests cloud services, the service provider accepts the request, creates the Virtual Machine (VM) as per the user's resource requirements and allocates it to the user. This process is called resource provisioning. Based on the requirements of cloud users, discovery and allocation of best workload – resource pair is an optimization problem (Singh & Chana, 2015). Resource provisioning is of two types based on the change in provision (Kukreja & Dalal, 2018). Static resource provisioning: It is suitable for the workloads whose requirements have no change over time. In such a case, the resources are provisioned for the users before actual execution. Dynamic resource provisioning: It is suitable for the workloads whose requirements change with time. Therefore, the resources are automatically scaled up or scaled down as per the customers' need even during the execution of the workload. Here, to deal with the dynamic nature of workloads the resources are reserved in advance or sometimes, the VMs are migrated among servers. Studies have found that servers in many existing data centers are often severely underutilized due to over provisioning for the peak demand (Armbrust et al., 2009). The overprovisioning problem can occur when the reserved resources for a certain customer exceed its demands, and without elasticity through non-peak-times, resources are wasted. In cloud computing, elasticity refers to the degree of automatic adaptation in resource provisioning in response to the continuous changes in the customer's workload and demand (Arani, Jabbehdari, & Pourmina, 2018). The under-provisioning problem can occur when the reserved resources are not suitable for the current customer's demands (Al-Ayyoub, Jararweh, Daraghmeh, & Althebyan, 2015).